What the best AI leaders get right about culture

Two companies buy the same AI licenses. In one, it spreads and changes how work gets done; in the other, it stalls and the seats go unused. The tools are identical, so the difference isn't the technology, it's the AI culture, specifically whether people feel safe to learn AI in the open instead of in secret. And a lot of it is happening in secret: a WalkMe survey found about 57% of US workers are reluctant to tell managers or colleagues they use AI, and Henley Business School found 48% hide their AI use for fear of being seen as lazy or incompetent. When half your workforce conceals the very tools meant to make them productive, you don't have a tooling problem, you have a culture problem, and it's the one thing a leader can actually move. The best AI leaders treat building that culture as the job, not a soft accompaniment to it.

The first thing they get right is noticing that the gap is a culture gap at all, which is harder than it sounds because leaders systematically misjudge how their own teams feel.

Leaders consistently assume the team is as far along as they are, and the data says they're badly off. Harvard Business Review, reporting BCG and Columbia Business School research, found 76% of executives believe their employees are enthusiastic about AI while only 31% of individual contributors actually are, leaders are more than two times off the mark. Gallup found the same shape in usage: in late 2025, 69% of leaders used AI versus 40% of individual contributors, and 19% of leaders were daily users against 11% of employees. So the executive looks at their own fluent, daily AI use, projects it onto the org, and concludes adoption is fine while the frontline has barely started. That misread is why mandates fall flat: they're aimed at a team the leader imagines, not the one that exists. Closing it starts with leaders seeing the real gap, which is the entry point to a serious AI for enterprise effort rather than a memo.

The move that separates leaders who build AI culture from those who suffocate it is the choice between enabling and surveilling, and it isn't a values preference, it's what the evidence shows works. Cornell research, reported through SHRM, found that algorithmic surveillance made workers perceive less autonomy and drove resistance, more complaining, worse performance, more intent to quit, but that the same AI produced far less resistance when it was framed as developmental feedback rather than monitoring. Infosys and MIT Technology Review found 83% of business leaders say psychological safety has a measurable impact on AI initiatives and 84% link it directly to business outcomes, yet only 39% rate their own org's psychological safety as high. The clarity piece matters here too: 60% said knowing how AI affects their job would most improve their sense of safety, so telling people what changes for their role, across roles and as part of an honest AI transformation, is itself a safety move. Whoever holds the Head of AI mandate usually owns keeping this culture honest.

The hard-nosed counter is that culture is soft and unmeasurable, that AI adoption is an execution problem, buy the licenses, mandate usage, track it on a dashboard, and adoption follows, and that talking about feelings is a way to dodge accountability for return. It sounds tough-minded, and it's exactly the move the data shows backfiring. Mandate-and-monitor is the approach that suppresses adoption: with around 57% of workers already hiding their AI use, a usage dashboard measures disclosure, not adoption, and surveillance further suppresses the behavior it's trying to count. Meanwhile the 'soft' variable is the one leaders themselves tie to hard outcomes, 84% link psychological safety directly to business results. So the choice isn't culture or execution; culture is the execution mechanism that decides whether your paid licenses get used in the open or used in secret. A leader's job isn't to deploy AI, it's to build a culture where people are safe to learn it, which is what turns a business full of unused seats into one where AI actually takes hold.